Texas sues OpenAI and wants court-approved checks on new models

Texas filed the case in Collin County district court. The petition says OpenAI marketed ChatGPT as safe. It says OpenAI failed to warn users. It cites the state's consumer protection law and its SCOPE Act for children. Texas asks the court to stop OpenAI from offering ChatGPT to minors in Texas. It also wants new models kept from Texas users until a court-approved evaluator signs off. It seeks civil penalties of $10,000 per violation, plus legal costs. OpenAI did not immediately respond to the Texas Tribune. These are allegations, and no court has ruled.

• Virginia issues Executive Directive 6. Governor Abigail Spanberger issued it on October 23. It orders a statewide plan against AI-related cyberattacks. It also creates a Cyber Consortium. Her office cited Anthropic AI agents that filed fake records requests on a state site. Its understanding is that no sensitive information was accessed. Sixty Bytes is made with Claude.
• Node.js patches a flaw an OpenAI tool reported. Node.js fixed 10 flaws on October 21. Four are rated high, in versions 26, 24 and 22. The worst, a Brotli decoder bug, can let attackers run code. Node.js credits OpenAI Codex Security with reporting it.
• Claude wrote machine-checked proofs for 123 libraries. Anthropic and Theorem published it on October 22. Claude wrote about 180 million lines of Rocq proofs for 239 entry points. They show no crashes on inputs under 64 MB, for RISC-V builds. The work found hundreds of bugs, and Anthropic notes that verified does not mean bug-free. Sixty Bytes is made with Claude.
• AI helped crack a 31-year-old geometry problem. Two researchers posted a proof of the KLS conjecture. It was proposed in 1995. The authors are Xinzhi Zhang of Microsoft Research and Zhao Song. Microsoft says AI agents explored approaches while the researchers chose the direction. The paper is an arXiv preprint, last revised October 4.